Fair Value Accounting and Firm Indebtedness Evidence from Business Combinations under Common Control

Transcription

1 Fair Value Accounting and Firm Indebtedness Evidence from Business Combinations under Common Control Massimiliano Bonacchi University of Naples Parthenope Antonio Marra SDA Bocconi School of Management Bocconi University Ron Shalev* Stern School of Business New York University This DRAFT: March 2015 *Corresponding Author: Ron Shalev Acknowledgements We are grateful to Andrea Toselli (EFRAG TEG Member and Technical Partner in PwC Italy) for helpful comments and suggestions on key issues of the topic. We also thank Elia Ferracuti and Palmira Piedepalumbo for excellent research assistance. We have received helpful comments and suggestions from Matthew Cedergren, Seil Kim, Pepa Kraft, Marco Mattei, Joshua Ronen, Paul Zarowin, and workshop participants at the University of Minnesota, NYU Stern, Tel Aviv University and the University of Padua. 1

2 Fair Value Accounting and Firm Indebtedness Evidence from Business Combinations under Common Control ABSTRACT: We analyze a choice that parent firms face under IFRS: whether to account a business combination under a common control (BCUCC) at fair value or at the historical cost, to provide evidence that firms would use fair value when they believe it would help them issuing public debt. A BCUCC is a merger of two entities owned by the same parent firm. Although most of BCUCCs do not materially change the composition and the market value of parent firm s assets and liabilities, they can significantly reduce accounting leverage of the parent firm if recorded at fair value. We find that parent firms are more likely to record BCUCCs at fair value when their pre-bcucc leverage is high and when they have net worth covenants on their debt. Using a propensity score to match firms that used fair value to account for a BCUCC with similar firms that did not conduct a BCUCC, we find that the former are more likely to issue new public debt following the BCUCC. Keywords: Accounting choices; fair value accounting; balance sheet leverage; indebtedness. 2

3 Fair Value Accounting and Firm Indebtedness Evidence from Business Combinations under Common Control I. INTRODUCTION Business combinations are among the most important corporate events. They are almost the only transactions through which firms can record intangible assets on the balance sheet. A unique form of business combination is a business combination under a common control (BCUCC). Paragraph B1 of IFRS 3 Business Combinations describes a business combination under common control as: a business combination in which all of the combining entities or businesses are ultimately controlled by the same party or parties both before and after the business combination, and that control is not transitory. BCUCCs are different from regular business combinations in two major aspects the transaction price and the transaction motivation may not reflect those of an arm s-length business combination. The IASB describe these differences as follows: (a) they are (BCUCCs) directed transactions rather than arm s-length exchanges and therefore the transaction price might not be representative of the fair value of the transferred business; and (b) the purpose of such transactions could be different from the purpose of business combinations that are not under common control. (IFRS staff paper: Business combination under common control, June 2014, page 1) Because of these unique characteristics, U.S. GAAP does not allow parent firms to use the acquisition method of accounting the method required for business combinations - for BCUCCs and requires parent firms to record these transactions at the historical cost carried on the parent firm balance sheet. Unlike US GAAP, IFRS 3 is silent on the accounting treatment for BCUCCs and allows parent firms to use a consistent policy to record such transactions (IAS 8.10). Parent firms choose between two methods: (1) the acquisition method, which is 3

4 consistent with fair value accounting and, (2) carrying the target s book values on the parent balance sheet, which is consistent with historical cost accounting. Under the acquisition method the parent firm allocate the purchase price to target s assets and liabilities. The purchase price allocation typically involves recording intangible assets and goodwill not previously recorded on the parent firm balance sheet and stepping up recorded net assets to the fair values. The likely effect of choosing the acquisition method over the historical cost to record a BCUCC is an increase in the book value of the parent firm and reduction in its accounting leverage. Extant literature in accounting provides evidence on the role of firm leverage in revaluation of fixed assets decisions. Easton, Eddey, and Harris (1993) survey Australian firm managers on the incentive to revalue tangible long-lived assets and suggest that the need to maintain low debt-to-equity ratio is an important factor in the decision to revalue long-lived assets. Aboody, Barth, and Kasznik (1999) analyze upward revaluations of fixed assets in U.K. companies and find that while the revaluations are associated with future performance, the association is much weaker in firms with high debt-to-equity ratio. The authors interpret the results as suggesting that when firms have an incentive to window dress their balance sheets, they tend to be opportunistic in applying the option to use fair value to step up their reported value of assets. In another study based in the U.K., Muller (1999) tests firms choice on whether to capitalize acquired brand names or to write them off immediately to equity. He tests a leverage hypothesis that firms will tend to capitalize brand names when leverage is high, but finds no support for the hypothesis. Finally, in a merger and acquisitions (M&A) setting, Aboody, Kasznik, and Williams (2000) provide evidence of a positive association between firms debt-to-equity ratio and the choice to use the purchase method, not the pooling of interest method to account for acquisitions. 4

5 A recent study by Christensen and Nikolaev (2013) provides a different perspective on the choice to use fair value accounting, in the context of firm leverage. The authors investigate firms choice, under certain IFRS standards, to commit either to fair value accounting or to historical cost accounting. Because IFRS objects to situations in which firms make fair value accounting choices on an ad-hoc basis and requires firms to develop a policy with regard to fair value option this choice is less likely opportunistic. The authors find that firms that rely more on debt financing tend to commit more to use fair value over historical cost accounting. The authors offer the following explanation: highly levered firms are required to provide fair value measurement to lenders as part of loan transactions and therefore find it less costly to reliably estimate, record, and report fair value of non-financial assets on financial statements. Acquisitions under common control provide us with a unique setting to extend the literature. Compared with a common business combination, in which the merger of the two entities may significantly affect the composition and values of acquiring firm s assets and liabilities, and thus also the expectation for future cash flows, a BCUCC is not likely to have a meaningful economic effect on the composition and value of parent firm asset and liabilities. We add to the above literature by studying whether firms opportunistically exploit the choice allowed by IFRS to record BCUCC at fair value in order to reduce accounting leverage in such a way that allows them to subsequently increase economic leverage without increasing their capacity to service debt. We identify a sample of European parent firms, of which two subsidiaries of the same parent went through a business combination and conduct the following analyses: 1 We start by validating our two working assumptions: (1) the choice to use the acquisition method for a BCUCC results in a decrease in balance sheet leverage. We find that firms that use the 1 We use the terms merger and acquisitions and business combination interchangeably. 5

6 acquisition method experience a large drop in accounting leverage immediately following the BCUCC (the ratio of debt to equity drops from 64.8% to 49.2% at the median level). A regression analysis suggests that the change in leverage is explained by the choice of accounting method to record the BCUCC; (2) BCUCCs, largely an internal decision, do not change, on average, the composition and market values of parent firm assets and liabilities and thus do not have a meaningful effect on parent firms expected future cash flows. We find that abnormal announcement-day returns are not statistically different from zero for both fair-value BCUCC parent firms and historical-cost BCUCC parent firms. We also find no difference in long-term stock performance based on the BCUCC accounting choice. Next we investigate the leverage hypothesis formalized in Muller (1999). Specifically, we test whether the likelihood of a parent firm of two merged subsidiaries to choose the acquisition method to account for the BCUCC rather than the historical cost increases with the parent firm s pre-bcucc leverage. We find that the likelihood of choosing the acquisition method increases with a firm s leverage. We also test whether the risk of violating net worth covenants drives as well the choice of a method to record the BCUCC. We focus on net worth covenants because a purchase price allocation in many cases results in the recording of a previously unrecorded intangible asset. Newly recorded intangible assets help a firm avoid a covenant violation only if these assets are not excluded from the covenant computation. Net worth covenants (unlike tangible net worth covenants) do not exclude intangible assets from the covenant computation. We find that firms with a net worth covenant on their debt are more likely to choose the acquisition method over the historical cost method. Finally, we test whether parent firms that conduct a fair value BCUCC attempt to take advantage of the window-dressed balance sheet in the public debt market by issuing new public 6

7 debt. Cohen and Zarowin (2010) document that firms window dress their earnings before seasonal equity offering, suggesting managers have incentive to window dress financial reports before public security offering. We focus on public debt because in Europe, bank debt is largely relationship-based (Boot 2000; Boot and Thakor 2000), and the landing bank likely possesses private information on a borrower. Thus, window dressing of the balance sheet, while possibly helping to avoid technical violations of existing debt covenants, is likely to be less effective in a bid to raise new bank debt. Public debt investors, however, are less likely than banks to fully undo the effect of the BCUCC on a firm s balance sheet leverage. Using a propensity scorematching (PSM) technique, we match firms based on the pre-bcucc balance sheet leverage and other firm characteristics and identify a control sample of firms that did not conduct a BCUCC. Analysis of the treatment and the control groups suggests that a treated firm is more likely than a control group firm to issue new bonds in each of the four quarters subsequent to the BCUCC. The marginal effect of a fair value BCUCC on the likelihood of issuing new public debt is of 9.6% at the first quarter and of 18.6% in the four quarters following the BCUCC. Taken together, our results suggest that firms make an accounting choice to use fair value in order to window dress their balance sheet when they are highly levered and will use the window-dressed balance sheet to increase their economic leverage without necessarily increasing their capacity to pay debt. Our study contributes to the existing literature in the following ways: First, it provides evidence that reported balance sheet values, holding the economic value of the net assets constant, may be a factor in firm s choice and likely also the ability to raise new public debt. Second, it highlights a unique transaction that under IFRS regime could result in a significant change on parent firms balance sheet and consequently indebtedness. With the SEC allowing 7

8 international firms listed in US exchanges to use IFRS this transaction can also impact US investors. Third, it suggests that even as IFRS requires a consistent method for similar transactions, firms may be opportunistic in making a fair value choice when the type of transaction necessitates a commitment that is less binding. Finally, it provides additional evidence on the effect of debt contracts on fair value accounting choice. Note that we do not argue and our evidence does not suggest that the accounting choice allowed is the main driver of BCUCCs. We do believe, however, that accounting choices are likely to be of higher order effect at BCUCCs than at arms-length business combinations. The remainder of the paper proceeds as follows. In Section 2, we review related literature and develop testable hypotheses. In Section 3, we discuss the sample selection process and descriptive statistics of the BCUCC population. In Section 4, we present our modeling and the related empirical results. We conclude in section 5. II. INSTITUTIONAL BACKGROUND, RELATED RESEARCH, AND HYPOTHESES DEVELOPMENT BCUCCs and IFRS A business combination under common control (BCUCC) is a transaction in which all of the combining entities or businesses are controlled by the same parent firm before and after the transaction. In terms of the parent firm financial reporting, both the acquirer and the target in the BCUCC were consolidated into the parent firm financial reports pre-bcucc and the merged firm continues to be consolidated into the parent firm reports post-bcucc. Figure 1 provides illustrations of possible structures of business combinations under common control. 8

9 BCUCCs represent a broad spectrum of transactions motivated by a range of different business purposes. A common reason provided by parent firms for conducting the BCUCC is rationalizing group operations, but other explanations such as, helping develop a premium brand at the acquiring firm, and to allow a superior balance between capital intensive (offshore) and less capital intensive (onshore) activities are also used. 2 In many cases the decision to engage in a BCUCC is an internal decision of the parent firm. As such both the transaction price and the transaction motivation can be different from what is observed in an arm-length business combination. For that reason, U.S. GAAP requires parent firms to record BCUCCs at historical cost and thus entails no change in firms financials as a result of the combination (ASC ). While U.S. GAAP and IFRS issued identical standards for business combinations conducted at arm s length between two transacting parties, IFRS allows parent firms to decide how to account for a BCUCC, provided they develop a system that will consistently apply the same accounting method to similar transactions over time. Unlike the fair value option in other IFRS standards, such as IAS 16 Property Plant and Equipment, and IAS 40 Investment Property, in which similar transactions occur frequently and hence a commitment to a consistent accounting method can be binding, BCUCCs are very infrequent and unique so that each such transaction practically involves an ad-hoc decision on the accounting method. IASB has long viewed the current state of BCUCCs accounting as undesirable. Thus in 2007 the board started a project on BCUCCs. The project was put on hold in 2009 and subsequently received high priority in The purpose of the project is to find commonalities in BCUCC transactions to enable the IASB to arrive at a consistent policy. To date no policy has emerged. Throughout our sample period, two methods are most often chosen to account for 2 Appendix B provides 4 examples of press releases in which the parent firms provide the motivation behind the BCUCC. 9

10 business combinations between entities under common control: the acquisition method, which entails evaluating a target s assets and liabilities at fair value through a purchase price allocation process; and the predecessor values method, which entails using the historical cost of a target s net assets that are recorded pre-bcucc on the parent balance sheet. In making the choice between the acquisition method and the historical cost method parent firms weigh the costs and benefits of using each method. The benefit of using fair values in the acquisition method is the reduction of balance sheet leverage. The costs of using acquisition method are increased likelihood of write-offs of goodwill, if it is recorded in the purchase price allocation, and potentially additional amortization and depreciation expense for newly recorded intangible assets and stepped up property plant and equipment. Li, Shroff, Venkataraman, and Zhang (2011) and Bens, Heltzer, and Segal (2011) show that goodwill impairment write-offs trigger significant negative market reactions. With regard to the additional depreciation and amortization, while Dechow, Huson, and Sloan (1994), and Gaver and Gaver (1998) find that CEO cash compensation is shielded from non-recurring losses, Shalev, Zhang, and Zhang (2013) suggest that managers would allocate the purchase price in a way that minimizes recurring items such as depreciation and amortization when they have compensation based incentive to do so. Accounting choices and debt contracts Fields, Lys, and Vincent (2001) define accounting choice as, Any decision whose primary purpose is to influence (either in form or substance) the output of the accounting system in a particular way. In the context of this study, the choice allowed by the IFRS to account for BCUCCs using the acquisition method effectively allows the parent firm to record multiple assets at fair value i.e., to step up the value of recorded assets and to record previously unrecorded 10

11 intangible assets on the balance sheet. The likely outcome of the fair value choice is an increase in the value of net assets on the balance sheet and a decrease in accounting leverage. The relation between debt contracts and firms accounting choices drew much attention of accounting literature. Beatty and Weber (2003) suggest that borrowers whose bank contracts allow accounting method changes to affect contract calculations are more likely to make income-increasing changes than income reducing changes. Beatty, Ramesh, and Weber (2002) provide evidence that borrowers are willing to pay an additional charge in order to retain flexibility in debt contracts with regard to accounting changes. The authors estimate that for borrowers, the extra cost for not excluding voluntary and mandatory accounting changes from the covenant calculation at 84 basis points for voluntary changes and 71 points for mandatory changes. The willingness to pay for accounting flexibility likely stems from the cost of covenants violations. Beneish and Press (1993) compile a sample of firms that went through a largeenough technical default and estimate the cost of renegotiating to refinance and restructure the debt at 0.37% of a firm s market equity value. Several studies document firms accounting choices firms make trying to circumvent debt covenants limitations. DeFond and Jiambalvo (1994) compile a sample of 94 firms that report a debt-covenant violation and show that these firms use accounting choices to increase earnings through accruals in the year leading, to the year of, the covenant violation. Sweeney (1994) compiles a sample of 130 firms that violated covenants and finds that firms respond with income-increasing accounting changes to the approaching default. The strength of the response is an increasing function of the cost of default and the flexibility allowed to the firm in the debt contract. 11

12 In the context of using fair values to value assets, academic literature generally suggests that reducing balance sheet leverage is one factor managers consider when they revalue assets at fair value. Easton, Eddey, and Harris (1993) survey Australian firm managers on the incentive to revalue tangible long-lived assets and suggest that the need to maintain low debt-to-equity ratio is an important factor in the decision to revalue long-lived assets. Aboody et al. (1999) analyze upward fair value revaluations of fixed assets in the U.K. and find a positive association with firm performance. This positive association, however, is weaker in the firms with high debt-to-equity ratio, suggesting that while in general firms use an upward fair value revaluation of fixed assets to convey positive information to investors, the incentive to window dress balance sheet leverage also plays a role in the decision. Courtenay and Cahan (2004) provide similar evidence based on New Zealand firms and capital markets. Further corroboration of the relation between revaluation of fixed assets and firms leverage is provided by Missonier-Piera (2007) who, focusing on Swiss firms, provides evidence on a positive association between revaluation of fixed assets and firms leverage. Finally in the specific context of business combinations, Aboody et al. (2000) analyze U.S. firms accounting choices to use either the purchase method (similar to the current acquisition method) or the pooling of interest method (essentially the historical cost) in business combinations. The authors use debt-to-equity ratio as a proxy for closeness to violating debt covenants and find that firms are more likely to choose the purchase method when the debt-to-equity ratio is high. Against the evidence of a positive association between firm leverage and the choice of fair value to revaluation of assets, Muller (1999) analyzes U.K. firms choice on whether to record purchased brand names on the balance sheet or to write them off immediately and finds no association between firms debt-to-equity ratio the choice to record brand names. 12

13 Finally, a recent study by Christensen and Nikolaev (2013) provide a different perspective on the relation between firm leverage and fair value accounting choice. The authors exploit the setting of IFRS fair value choice, in which under IAS 16 Property Plant and Equipment and IAS 40 Investment Property, a firm can use either fair value or historical cost to account for certain long-lived assets and investments. Firm are required, however, that the choice not be ad-hoc, rather it should be consistently applied to similar transactions over time. The requirement from managers to commit to sticking with one accounting method makes it less likely that their choice is subject to managerial opportunism. The authors find that firms will generally be more inclined to commit to use fair values for more liquid assets. In the context of debt financing, the authors find that firms that rely more heavily on debt financing are more likely to take the fair value choice. The authors explain their finding with borrowers, having to provide lenders with fair value measurement of their assets to secure the debt, find it less costly to reliably estimate and report the fair value of non-financial assets on financial reports. Though the general IFRS requirement of fair value choice to be consistent over time holds for BCUCCs, in practice the choice made on a BCUCC is largely ad hoc. As reorganizations which give rise to BCUCCs are very infrequent and the motivation for each BCUCC could be different from the previous one, distinguishing one from another is relatively easy, which make the commitment for a consistent method less binding. As most empirical evidence points to firms exploiting opportunities to use fair value to increase the value of assets on the balance sheet when the balance sheet leverage is high, we formalize our first testable hypothesis as follows: 13

14 H1a: The likelihood of a parent firm to record a BCUCC using the acquisition method rather than the historical cost method increases with firm leverage. We also hypothesize that the costs of technical violations of debt covenants may drive firms to make accounting choices that can avert the risk of violating a covenant. The covenant that is most obviously affected by the accounting choice for a BCUCC is a net worth covenant. Whereas intangible assets are excluded from the covenant computations of tangible net worth covenants, thus neutralizing the effect of any increase in intangible assets through the purchase price allocation, they are not excluded from net worth covenants. Therefore, the follow-up to the first hypothesis is: H1b: The likelihood of a parent firm to record a BCUCC using the acquisition method rather than the historical cost method increases in firms with net worth covenants on their loans. The underlined conjecture in H1a and H1b is that firms would want to window dress their balance sheet when incentivized to do so by debt contracts. A natural follow up inquiry is whether firms take advantage of the window dressed balance sheet beyond avoiding current debt covenants. Specifically, we are interested in whether firms make use of the windowdressed balance sheet to issue new public debt and thus increase economic indebtedness without fundamentally increasing their ability to service the debt. Cohen and Zarowin (2010) document that managers window-dress their firms earnings before seasonal equity offering, suggesting that managers have an incentive to window dress financial reports before a public security offerings. We focus on public debt issuance because in Europe, bank loans are typically relationship-based (Boot, 2000; Boot and Thakor, 2000). As such the quality of information banks possess on borrowers is relatively high, which increases the likelihood of banks undoing 14

15 the effect of the fair-value BCUCC on firms balance sheet. Under relationship banking, balance sheet window dressing may help firms avoid technical violations of current net worth covenants, but likely to prove less effective in raising new bank debt. With public debt, by contrast where information asymmetries are greater and monitoring is relatively weaker than bank debt, window dressing of balance sheet leverage may prove more effective. Therefore, we predict that we are likely to observe more bond issuance from firms that have gone through a fair-value BCUCC relative to comparable firms that did not go through a BCUCC. Hence our second testable hypothesis is as follows: H2: Parent firms that used the acquisition method to record a BCUCC are more likely to issue new public debt in the period following the BCUCC than comparable firms who did not perform a BCUCC. III. SAMPLE SELECTION AND DESCRIPTIVE STATISTICS Sample selection To identify business combinations under common control (BCUCC), we start with SDC database and select acquisitions that satisfy the following requirements: (1) the acquirer gained control over the target in the transaction; (2) acquirer and target both have the same immediate or ultimate parent company, and (3) the parent firm is incorporated and headquartered within the European Union nations. Our sample begins in year 2005, the first year in which IFRS was mandatory for the consolidated financial statements of all listed firms in Europe. 3 We drop the sample transactions in which the acquirer, seller, or parent is a financial institution. To avoid confounding effects, both in the economics of the BCUCC and the accounting classification to 3 Christensen, Hail, and Leuz (2013) provide a list of countries shown in Table 1 with relative dates when IFRS reporting becomes mandatory. 15

16 acquisition method or historical cost, we drop transactions in which the parent (or group) was involved in more than one acquisition during the reporting period. We require parent firms in the sample to have financial and stock price data available from Compustat Global. After employing the above restrictions there are 421 business combinations under common control. Next, we classify the BCUCCs to acquisition method (fair value) and historical cost. We use the change in goodwill at the parent firm to initially identify whether the acquisition method was used to account for the BCUCC. Transactions in which the parent firm reports an increase in goodwill in the BCUCC quarter, 147 in total, are classified as acquisition method. We verify were information is available that these transactions were accounted using the acquisition method. Transactions in which the parent firm reports no change in goodwill, 83 in total, are classified as historical cost. The remaining transactions are such that goodwill on the parent firm balance sheet decreased in the BCUCC quarter. Goodwill impairment on the parent firm balance sheet can be related to the BCUCC, if the transaction price in the BCUCC is lower than the carrying value of the target net assets on the parent balance sheet, which may occur since the price not necessarily reflect the market price of the target. In that case, both accounting methods would lead to a similar outcome. The goodwill impairment could also be unrelated to the BCUCC. In either case, there is no effective way to identify the methods used to account for the BCUCC. Therefore, we eliminate these transactions from the sample. The final number of observations used in the analysis is 230. Table 1 reports the sample selection process in detail. Descriptive statistics Table 2, panel A reports a sample breakdown for fair-value BCUCCs and historical-cost BCUCCs by fiscal year. While the portion of firms within a sample year that choose the acquisition method to account for a BCUCC ranges from 41.94% in 2012 to 79.31% in 2009, 16

17 there is no clear clustering across specific years. Table 2, panel B reports industry distribution using the Fama-French 12-industry classification. The portion of BCUCCs with the fair value accounting choice ranges from 50% in Energy, Oil, Gas, and Coal Extraction to a high of 84.62% in Chemicals and Allied Products with no clear clustering across industries. Finally, panel C reports a sample distribution by country of incorporation. The number of BCUCCs ranges from 1 in Luxemburg to 25 in Germany. Table 3 reports descriptive statistics of the sample firms, broken down by BCUCC accounting choice. There is no statistically significant difference between the two groups in terms of parent firm size, book value, and market-to-book ratio. Firms that use fair value are exante more profitable in terms of mean return on assets (ROA), but there is no significant difference for the median ROA. In addition, 39% of the BCUCC accounted for at fair value had minority interest involved in the transaction compared with 32.5% percent of BCUCCs accounted for at historical cost. Thus, there is no significant difference between the two groups in the proportion of BCUCCs in which the parent firm does not fully own the acquirer, the target, or the immediate parent of the target (MINORITY). With regards to the variables of interest, firms that choose the acquisition method to account for a BCUCC are ex-ante more levered with mean (median) pre-combination debt-to-equity ratio (D/E_pre) of 67.2% (64.8) compared with 54.1% (47.0) for firms that elected to account for a BCUCC using the historical cost of target net assets. The difference between the means and the medians is significant at the 1% level. Measuring leverage as the debt-to-assets ratio instead of debt-to-equity yields a similar relation. Further, the post-bcucc debt-to-equity ratio exhibits a sharp decline for parent firms that accounted for the BCUCC using fair value when compared to the pre-bcucc debt-to-equity ratio with a mean of 58.7% (compared with 67.2 pre-bcucc) and median of 17

18 49.02% (compared with 64.8% pre-bcucc). Firms that elected to account for the BCUCC using the target s historical cost of net assets do not exhibit a similar decline. Finally, a larger portion of the firms that use fair value to account for the BCUCC had a net worth covenant associated with their bank debt (38.1%) than the firms that use historical cost (24.1%). The statistics provide preliminary evidence in support of the leverage hypothesis and in line with our predictions in H1. IV. RESEARCH DESIGN AND EMPIRICAL FINDINGS Validity Tests Two assumptions underlie the analyses conducted in this study. The first assumption is that accounting for BCUCCs using the acquisition method reduces parent firms balance-sheet leverage and that the reduction is larger than for a BCUCC accounted at the historical cost method. The second assumption is that a BCUCC does not materially change parent firms assets and liabilities composition and their market value and thus is not expected to affect future cash flows. In this section we validate both assumptions. BCUCC Reduces Accounting Leverage We validate our first maintained assumption that the choice to account for a BCUCC using the acquisition method rather than the historical cost indeed results in a reduction in the parent firm balance-sheet leverage. It is not ex-ante certain the fair value choice would lead to a lower accounting leverage. A fair-value BCUCC involves an assessment of a group of assets and liabilities at fair value. When, for example, the value of target debt on a parent s balance sheet was previously recorded at values significantly lower than the fair value or when the parent firm delayed recording write downs of its assets, the balance sheet leverage could 18

19 actually increase. If parent firms really have reducing leverage in mind when they make the accounting choice of fair value to record BCUCC, this choice should actually result in lower accounting leverage post-bcucc. To test this we estimate the following OLS regression: LEVERAGE_CH i = β 0 + β 1 FV_BCUCC i + β 2 7 CONTROLS i,t 1 + ε i, (1) where LEVERAGE_CH is the change in leverage measured alternatively as the change in debtto-equity ratio between post-bcucc and pre-bcucc (post-bcucc debt-to-equity ratio minus pre-bcucc debt-to-equity ratio) and the change in debt-to-assets ratio between post-bcucc and pre-bcucc. FV_BCUCC is an indicator variable that takes the value of 1 for a BCUCC carried out using the acquisition method and zero otherwise. We include the same control variables that are included and described in detail following equation 3. 4 Results reported in Table 4 validate our maintained assumption that the change in leverage is negatively associated with the choice to account a BCUCC using the acquisition method (debt-to-equity ratio: coefficient= ; z-stat= -2.63; debt-to-assets ratio: coefficient= ; z-stat= -2.87). Economically, the choice of fair value to account for a BCUCC results in an average drop of 14.1% in firms debt-to-equity ratio and in a 6.8% drop in the debt-to-assets ratio. The results suggest that the choice to account for a BCUCC using fair value results in a significant drop in the parent firm accounting leverage. BCUCC Accounting Choice is not Value Relevant Extant accounting literature suggests that fair value accounting is value relevant (See Barth, Beaver and Landsman, 2001, for a review of the literature). Specific to asset revaluation, Aboody et al. (1999) find that asset revaluations are informative about future firm performance. 4 We only exclude the level of goodwill from the control variables. The inclusion of this variable does not change results. 19

20 Because BCUCC are conducted for various reasons, an argument could be made that in the choice of whether to use fair value or historical cost to record a BCUCC, managers convey their private information of whether the BCUCC is expected to create value or not. To alleviate this concern, we conduct several return analyses that compare fair-value choice parent firms and historical-cost choice parent firms. Specifically, we perform the following analyses: (1) We compare the average announcement day returns 5 of the two BCUCC groups. We find no statistically significant difference between them. The mean (median) market adjusted return is 0.1% (0.3%) for the fair-value BCUCC and 0.5% (0.3%) for the historical cost BCUCC. (2) We follow Aboody et al. (1999) and test whether there are significant differences in the longterm stock performance between the fair-value BCUCC parent firms and the historical-cost BCUCC parent firms. Specifically we estimate the following regression: RETURN i = β 0 + β 1 FV_BCUCC i + β 2 EBITDA i + β 3 EBITDA_CH i + ε i. (2) We measure the variables in this analysis following Aboody et al. (1999). RETURN i is measured as a firm s stock return in the year starting six months before the BCUCC announcement. FV_BCUCC i is an indicator variable that takes the value of 1 for a BCUCC carried out using the acquisition method (fair value) and zero otherwise. EBITDA i is parent firm earnings before depreciation, amortization and interest deflated by the firm market value of equity in the year starting six months before the BCUCC announcement, and EBITDA_CH i is the change in EBITDA i from a year leading to the BCUCC to the year following the BCUCC. We control for industry, country, and year fixed effects. Table 5 reports the results. We compare fair-value BCUCC parent firms performance to both the Compustat Global universe and to historical-cost parent firms. Column 1 reports the 5 Results untabulated. 20

21 results for a sample that includes the entire Compustat Global universe. Column 2 reports results only for the firms that performed a BCUCC. Both analyses provide consistent evidence of no statistically significant difference in the long-term returns between firms that performed a fair-value BCUCC and both firms that performed a historical-cost BCUCC and the entire universe of Compustat Global firms. In columns 3 and 4 we repeat the analysis of columns 1 and 2 but measure returns following the Fama-French (1993) three-factor model abnormal returns. Results are qualitatively similar. The effect on balance sheet leverage on the choice of accounting method To test the predictions in H1a and H1b that the likelihood of a choice to account for a BCUCC using the acquisition method increases with the parent firm s pre-bcucc leverage and with the existence of net worth covenants in the parent company bank debt, we estimate the following logistic regression: FV_BCUCC i = β 0 + β 1 LEVERAGE_pre i + β 2 COVENANT i + β 3 _ 9 CONTROLS i,t 1 + ε i, (3) where FV_BCUCC is an indicator variable that takes the value of 1 for a BCUCC carried out using the acquisition method and zero otherwise. The independent variables in the regression are measured when applicable at the quarter prior to the BCUCC. The variables of interest are the following: LEVERAGE_pre is a parent firm s leverage before the BCUCC. It is measured alternatively as the parent firm s total book value of debt scaled by the book value of equity or as the parent firm s total book value of debt scaled by total assets. COVENANT is an indicator variable that takes the value of 1 if the parent firm discloses a net worth covenant in the financial statements footnotes and zero otherwise. 6 6 Information on the presence of a net worth debt covenant is handcollected from the listed parent financial statements in the quarter before the BCUCC took place. 21

22 We include the following control variables in the analysis: firm size (SIZE), firm market-to-book ratio (MTB), the BCUCC method of payment (CASH), the level of goodwill at the parent firms (GDW), parent firm performance (ROA), the change in the level capital expenditure at the parent firm from before to after the BCUCC (CAPEX_CH), and whether the parent company own 100% of the shares in both merging subsidiaries (MINORITY). SIZE is measured as the natural logarithm of the parent firm total assets. MTB is measured as the ratio of the firm market value of equity to its book value of equity. CASH is measured as the cash percentage of the total consideration paid. We include this variable as the acquisition financing that may affect the choice of accounting method. GDW is measured as the goodwill on the parent firm s balance sheet scaled by the total assets. As one of the costs of the choice to use the acquisition method for BCUCC is an increased likelihood of future impairment of goodwill, the pre-bcucc level of goodwill on the balance sheet of the parent firm may affect its willingness to record more goodwill and thus also the incentive to use the acquisition method. ROA is measured as income before extraordinary items scaled by total assets, which is a commonly used measure to control for firm performance. Parent firm performance may affect the need to issue more debt as well as the probability of avoiding covenant violations and thus affect the choice of accounting for the BCUCC. CAPEX_CH is measured as the difference between averages over four fiscal quarters of post-bcucc quarterly-cash flow statement reportedcapital expenditure and the pre-bcucc capital expenditure. We control for change in the capital expenditure to account for firms need of cash for investment. Greater need for cash is likely to be positively associated with the need to window dress the balance sheet in order to help in raising debt. MINORITY is an indicator variable that takes the value 1 if the parent firm is not a sole owner throughout the chain of ownership in both the acquirer and the target in the 22

23 BCUCC and zero otherwise. We control for the existence of a minority interest in one of the transaction parties because such an existence could affect the transaction terms and ultimately the accounting choice for the BCUCC. In all analyses we include year, industry, and country of incorporation fixed effects. Results are reported in Table 6. Column 1 reports results for pre-bcucc debt-to-equity ratio as the explanatory variable, column 2 reports results for pre-bcucc debt-to-total assets ratio as the explanatory variable, column 3 reports results for the existence of a net worth covenant as the explanatory variable and columns 4 and 5 report results for regressions that include both explanatory variables (where debt-to-equity and debt-to-assets alternatively measure leverage). Regression analyses provide results consistent with the empirical predictions in H1a and H1b. Both coefficients on LEVERAGE_ pre (debt-to-equity ratio: coefficient=3.509, z-stat=2.87, debt-to-assets ratio: coefficient=1.368, z-stat=2.51) and COVENANT (column 3: coefficient=1.026, z-stat=2.29) are positive and significant. These results suggest that when making the choice of whether to use the acquisition method or the historical cost of a target s net assets, managers consider the balance sheet leverage and whether the choice would affect the likelihood of violating debt covenants. Fair-value BCUCC and the likelihood of raising new public debt In this section we test whether parent firms that engaged in fair value BCUCCs actually attempt to take advantage of the decrease in balance sheet leverage and issue more public debt. The purpose of this analysis is to investigate whether the accounting choice for a BCUCC can have a real measurable effect in the form of an increase in firms economic indebtedness without a clear increase in firms ability to serve debt. Because historical cost BCUCC firms are not necessarily similar to fair-value BCUCC in respect to the ex-ante likelihood of issuing new 23

24 public debt, we match our sample of fair-value BCUCC, with a sample of firms with similar pre-bcucc characteristics. For each parent firm that conducted a fair-value BCUCC we identify a matched firm based on characteristics described below. We then pool the two groups of firms and perform the analysis. To identify a matching firm to each fair-value BCUCC firm ( a treatment firm) we apply a propensity score-matching (PSM) 7 procedure developed by Rosenbaum and Rubin (1983), extended by Heckman, Ichimura, and Todd (1997), and introduced to the accounting literature by Armstrong, Jagolinzer, and Larcker (2010). Specifically, we estimate the following logit regression: FV_BCUCC i = β 0 + β 1 LEVERAGE_pre i + β 2 SIZE i + β 3 MTB i + β 4 CAPEX_CH i + β 5 RESEARCH i + β 6 EBITDA_CH i + ε i, (4) where FV_BCUCC is an indicator variable equal to 1 if a parent firm accounted for a BCUCC using the acquisition method and zero otherwise. LEVERAGE_pre is a firm s leverage before the BCUCC. It is measured alternatively as the parent firm s total book value of debt scaled by the book value of equity or as the parent firm s total book value of debt deflated by total assets. SIZE is the natural logarithm of the parent firm s total assets. MTB is the ratio of the firm s market value of equity to its book value of equity. CAPEX_CH is the change in capital expenditure from pre-bcucc to post-bcucc. RESEARCH is the ratio of research expense to sales at the parent company. Research expense gives rise to unrecognized intangible assets off the firm s balance sheet. These assets are likely to be recorded if fair value is chosen. The ability to record previously unrecorded intangible assets may affect the BCUCC accounting choice. Ideally, research should be measured at the acquisition target-firm level. Since financial data on subsidiaries is not available, we use data at the parent-firm level. EBITDA_CH i is the 7 We use a PSM to better match fair-value BCUCC parent firms with comparable firms. All results hold if we compare fair value BCUCC with historical cost BCUCC firms. 24

25 change in operating income before depreciation and amortization deflated by the parent market value of equity. We include this variable as a proxy for parent firm performance, in order to account for the possibility that despite our described above validity test results, evidence on the value relevance of fair value revaluation (Aboody et al. 1999) holds for our treatment firms. 8 We also include industry, country, and year fixed effects. We use a one-to-one nearest-neighbor matching without replacement (Heckman et al. 1997), restricting the attention to a falling propensity score in the common support for both groups (Smith and Todd 2005). 9 Using the predicted probabilities propensity scores from the logistic regression, we then match each fair-value BCUCC observation with the observation from the control group, which minimizes the absolute value of the difference between propensity scores. 10 In order to avoid matched pairs with significant differences in the propensity score, we also impose a tolerance level on the maximum propensity score distance smaller than 0.5% (caliper). Table 7 reports the propensityscore estimation results based on a pool of 150,329 observations. 11 Panel A of Table 7 reports results of the PSM regression and panel B of Table 7 reports descriptive statistics of the treatment firms and control firms with respect to matching variables. Reported statistics suggest that the matching process results in a control group of firms that is very similar to the treatment group in all the important respects (i.e., size, leverage, fixed-asset investment plans, research expense, changes in EBITDA, industry, country, and year). 8 Results are qualitatively similar if we do not match based on this variable. 9 The common support condition drops observations in which the propensity score is smaller than the minimum and larger than the maximum in the opposite group. This restriction rules out the phenomenon of perfect predictability; i.e., it ensure that firms with the same X values have positive probabilities of being both treated or not. 10 We also use additional PSM modeling, with unchanged results (see for details sensitivity and robustness check section). 11 We report results only for the debt-to-equity measure of leverage. Results are qualitativly similar when we use debt to assets. 25

26 We then pool the matched firms produced by the PSM process with the firms that conduct fair-value BCUCC and test whether the likelihood of issuing new public debt is different between the two groups. H2 predicts that firms that conducted fair-value BCUCCs are more likely than the matched firms to issue new public debt in the period shortly following the BCUCC. We start with a univariate comparison of a proportion of firms that issued new bonds in the four fiscal quarters immediately following the BCUCC between the two groups the fairvalue BCUCC firms and the subsample of matched firms. Because issuing new debt could also serve to replace old debt that was paid, without effectively increasing indebtedness, we require that a firm s debt level increase following the new debt issuance. Table 8, panel A reports results. In the quarter immediately following the BCUCC, 13% of the fair-value BCUCC firms issued new public debt compared with 4.1% of the control firms. The difference persists for three additional quarters in which the cumulative proportion of firms issuing new public debt is 21.9% for fair-value BCUCCs and 7.5% for the matched sample. All differences are statistically significant at the 1% level. These differences suggest that firms that conduct fair-value BCUCCs are more likely than similar firms that did not conduct fair-value BCUCCs to issue new bonds in the periods following the BCUCC. Next, we use a regression analysis to test whether controlling for additional factors can change the inference drawn from the univariate analysis. To that end, we estimate the following logistic regression: ISSUE i,t+1_t+4 = β 0 + β 1 FV_BCUCC i + β 2 CONTROLS i,t 1 + ε i, (5) where ISSUE is an indicator variable that takes the value of 1 if a sample firm satisfies the two following conditions: (1) The firm issues new debt in the four quarters post-bcucc, and (2) the levels of debt on the firm s balance sheet increased following the debt issuance and zero 26

27 otherwise. FV_BCUCC is an indicator variable that takes the value of 1 for a parent firm that recorded a BCUCC using the acquisition method and zero otherwise. We include the following control variables: SIZE is the natural logarithm of a firm s total assets. LEVERAGE_pre is a firm s leverage before the BCUCC. CAPEX_CH is the change in capital expenditure from pre- BCUCC to post-bcucc. RESEARCH is the ratio of research expense to sales at the parent company. EBITDA_CH i is the change in operating income before depreciation and amortization deflated by parent market value of equity. Results are reported in Table 8, panel B. Column 1 reports results for a new bond issuance in the first quarter following the BCUCC and columns 2 (3 and 4) reports results for the first two (three and four) quarters following the BCUCC. Consistent with H2, as well as with evidence from the univariate analysis, firms that engage in fair-value BCUCCs are more likely to issue new public debt following the BCUCC. The coefficient on FV_BCUCC is positive and significant at the 1% level across all regressions. Economically, the effect of a fair value BCUCC is no trivial. The marginal effect of a fair value BCUCC on the likelihood of issuing new public debt is of 9.6% at the first quarter following the BCUCC and of 18.6% in the four quarters following the BCUCC. These results suggest that firm managers perceive lower balance sheet leverage advantageous in public debt issuance and attempt to take advantage of the face lifted balance sheet and issue new public debt following a fair-value BCUCC. This leads firms to increase their indebtedness without a real change in their capacity to service the debt. Sensitivity tests and robustness checks In principle, one cannot rule out the possibility that our results are not the outcome of a firm s accounting choice to use fair values to account for the BCUCC but of the BCUCC proper. For example, one may argue that a BCUCC makes the structure of a conglomerate more 27